Gallup, Mason-Dixon, New York Times -Sienna, Quinnipiac University and other polls are facing competition from a newcomer this election season — the prediction markets.

Polymarket, Kalshi and Predictit are some of the prediction market companies raking in millions of dollars from midterm election bettors. On Election Night, political scientists will be watching how the prediction markets compare to polling.

Prediction markets are a new phenomenon. They’ve gained a mainstream following, with CNN, the Wall Street Journal and Fox News among the networks who are partnering with prediction market companies.

Prediction markets work like this: A person can put a bet down on anything from a sports contest to an election to what country will reach Mars first. The bettor buys contracts. For example, on Friday night 68 percent of the people placing bets in the Florida governor’s race picked Byron Donalds to win. The bettors will receive one dollar for each contract they purchase, a 32-cent profit per contract. The prices change with each purchase, like a stock price changes.

This will be the first midterm election since the markets have exploded in popularity. By early in October $1.4 billion in midterm bets had been made and by Election Day bets are expected to exceed $2 billion, according to the Betting for Democracy report. The amount spent far exceeds the $92 million bet on the 2024 presidential election.

The accuracy of prediction markets caught the attention of professors at Vanderbilt University in Nashville. The first study was motivated in part by inaccuracies in polling data from several past presidential elections.

“Our curiosity was originally are markets more accurate than polls? But the abstracted question is, can prediction markets tell us things about the world earlier than traditional methods? And that is what we are continuing to collaborate on to see if that is a useful tool, primarily in the national security space,” said Brett Goldstein, who works for the school’s Institute of National Security.

The study done by Goldstein, Laurie Cutting, Sarah Hughes-Berheim and others did find that besides being more accurate than some polls the prediction markets predicted Trump was the winner earlier than the polls.

Their study, using Polymarket, found that by midnight of election night 95 percent of the money was saying Trump was the winner. The Associated Press didn’t call the election for Trump until several hours later.

The prediction markets were showing Trump winning, as early as October, while polls were giving Kamela Harris the edge or calling it a toss-up.

“What we found that was as early as mid-October the Polymarket data was suggesting Donald Trump would win the election while the polls at that point were still unclear,” said Hughes-Berheim. “And what we were also really excited to see or interested to see is that this was also the case for five out of the seven swing states, which those again showing pretty strong indications that Donald Trump would win as early as mid to late October while the polls were still showing like the 50-50 boundary and they were unclear.”

Joshua Clinton and TzuFeng Huang, also from Vanderbilt, did a study specifically looking at how accurate the prediction markets were in the 2024 presidential race.

They looked at more than 2,500 bets and $2 billion in transactions during the final five weeks of the election. They found that Kalshi correctly predicted 78 percent of the outcomes on Election Day eve. Traditional polls only were 63 percent accurate, according to their study.

Why polls and prediction markets vary?

“Unlike polls which reflect the opinions of voters at the time of the survey, prediction markets are intentionally forward-looking and backed by financial stakes of the traders involved,” Clinton wrote in his paper.

There’s another reason polls and prediction markets differ, Cutting said.

Polls ask who the person is going to vote for; prediction markets ask who they think will win, she said.

FGCU professor Sean Ewing prefers polls over the prediction markets.

“A poll that's done right is a more accurate snapshot of the American public,” he said.

Polls and prediction markets tend to be closely correlated because he thinks polls influence how people will bet.

“It's a good indicator and it's easier than polls because it’s a bigger average, whereas polls are sort of individual snapshots at one moment in time. But I think that they're often informed by the polls, so it's almost like a secondary way to view a particular race,” Ewing said.

POLLS VS. PREDICTION MARKETS

Aggregate polls by FiftyPlusOne except when noted and percentages are from bets placed on Kalshi as of Saturday, October 31

Florida Governor Race

Kalshi
Byron Donalds 64%
David Jolly 36%
FiftyPlusOne Average
Donalds +5 points

Florida Senate Race

Kalshi
Ashley Moody 88%
Angie Nixon 12%
FiftyPlusOne Average
Ashley Mood y +7 points

Florida House District 22

Kalshi
Pia Dandiya 72%
Casey Askar 28%
Redbox Research +1

Florida House District 19

Kalshi
Jim Schwartzel 96.7%
Victor Arias 3.3%
No public polls available

Texas Senate Race

Kalshi
James Talarico 71%
Ken Paxton 29%
FiftyPlusOne Average
Talarico +3

Ohio Senate Race

Kalshi
Sherrod Brown 71%
Jon Husted 29%
FiftyPlusOne Average
Sharrod Brown +3

Ohio Governor’s Race

Kalshi
Amy Acton 71%
Vivek Ramaswamy 29%
FiftyPlusOne Average
Acton +3

Michigan Senate Race

Kalshi
Abdul El-Sayed 77%
Mike Rogers 23%
FiftyPlusOne Average
El-Sayed +5

Maine Senate Race

Kalshi
Troy Jackson 69%
Susan Collins 31%
FiftyPlusOne Average
Jackson +4

Kansas Senate Race

Kalshi
Adam Hamilton 51%
Roger Marshal 49%
FiftyPlusOne Average
Hamilton +3

WGCU is your trusted source for news and information in Southwest Florida. We are a nonprofit public service, and your support is more critical than ever. Keep public media strong and donate now. Thank you.